Ommi

Caution

Ommi is under construction and much of the functionality is undergoing frequent revision. There is no guarantee future versions will be backwards compatible.

Have you ever needed to use a database for a simple project but didn't want to worry about which database you were going to use? Or maybe you wanted to create a package that needed to store data but didn't want to force your users to use a specific database? Meet Ommi, a simple object model mapper that allows you to use whatever models you want to interface with whatever database you like.

Ommi doesn't provide its own model types, it allows you to use whatever models you are already using. Compatibility with the most popular model implementations are ensured through a comprehensive unit test suite.

Compatible Model Implementations

Ommi's test suite checks for compatibility with the following model implementations:

• Dataclasses
• Attrs
• Pydantic

Included Database Support

SQLite3

• Table creation from models
• Select, Insert, Update, Delete, Count
• One-to-one & one-to-many relationships
• Joins to filter queries for a single model type (cannot query for multiple model types in a single query)

PostgreSQL

• Table creation from models
• Select, Insert, Update, Delete, Count
• One-to-one & one-to-many relationships
• Joins to filter queries for a single model type (cannot query for multiple model types in a single query)

MongoDB

• Collection creation from models
• Fetch, Insert, Update, Delete, Count
• One-to-one & one-to-many relationships
• Joins to filter queries for a single model type (cannot query for multiple model types in a single query)

Usage

Defining Models

All models that support Ommi database drivers need to use the ommi_model class decorator.

from ommi import ommi_model, Key
from dataclasses import dataclass
from typing import Annotated

@ommi_model
@dataclass
class User:
    name: str
    age: int
    id: Annotated[int, Key] = None  # Optional primary key

Models can be assigned to model collections. Any model not assigned a collection will be assigned to a global collection which can be accessed by calling ommi.models.get_global_collection().

from ommi.models.collections import ModelCollection

collection = ModelCollection()


@ommi_model(collection=collection)
@dataclass
class User:
    name: str
    age: int

Connecting

from ommi.ext.drivers.sqlite import SQLiteDriver, SQLiteConfig


async def example():
    async with SQLiteDriver.from_config(SQLiteConfig(filename=":memory:")) as db:
        ...

Database Actions

The database drivers provide add, count, delete, fetch, sync_schema, and update methods. These methods should be wrapped in an ommi.drivers.DatabaseAction. The database action will capture the return and wrap it in a DatabaseStatus result that is either a Success or Exception. The database action provides an or_raise method that will force the exception to be raised immediately or return a Success result. The DatabaseStatus types are sub-types of tramp.results.Result types.

Add

Add takes any number of model instances and adds them to the database.

user_1 = User(name="Alice", age=25)
user_2 = User(name="Alice", age=25)
await db.add(user_1, user_2).raise_on_errors()

Find

Find returns a FindAction. This action is used to count, delete, fetch, or set fields on models in the database. It takes any number of predicates that are AND'd together. It doesn't return any models or make any changes to the database on its own.

Count

Count is an action of find that returns the number of models that match the predicates passed to find. It returns an AsyncResultWrapper which allows you to access the returned int value through chaining, you can read more about it below.

count = await db.find(User.name == "Alice").count().value

Delete

Delete is an action of find that deletes all models that match the predicates passed to find. It also returns an AsyncResultWrapper.

await db.find(User.id == user.id).delete().raise_on_errors()

Fetch

Fetch is an action of find that returns all models that match the predicates passed to find. It provides all and one helper methods to help with value unpacking, they both raise on errors. Calling fetch directly will return an AsyncResultWrapper that contains the list of models.

users = await db.find(User.name == "Alice").fetch().value

users = await db.find(User.name == "Alice").fetch.all()

user = await db.find(User.name == "Alice").fetch.one()

Models provide a reload method that will pull the latest data from the database. It returns an AsyncResultWrapper.

await user.reload().raise_on_errors()

Set

Set is an action of find that updates all models that match the predicates passed to find. It takes a any number of keyword arguments that are used to update the models fields. It returns an AsyncResultWrapper.

await db.find(name="bob").set(name = "Bob").raise_on_errors()

Models provide a save method that will push changed fields to the database. It returns an AsyncResultWrapper.

user.name = "Bob"
await user.save().raise_on_errors()

Schema

Schema is an action object that provides methods to manipulate the database itself. It can optionally be passed an optional ModelCollection or nothing to default to the global collection.

Its create_models action returns an AsyncResultWrapper that contains a list of the models that were created.

await db.schema().create_models().raise_on_errors()

await db.schema(some_model_collection).create_models().raise_on_errors()

It also provides a delete_models action that deletes all models in the collection from the database.

await db.schema(some_model_collection).delete_models().raise_on_errors()

AsyncResultWrapper

AsyncResultWrapper is a wrapper around the result of an async database action. It provides various awaitable properties and methods that allow you to access the result of the action. Awaiting the AsyncResultWrapper itself will return a DatabaseResult.Success object if the action succeeded or a DatabaseResult.Failure object if there was an exception.

match await db.find(User.name == "Alice").count():
    case DatabaseResult.Success(value):
        print(value)

    case DatabaseResult.Failure(error):
        print(error)

Value

Value is an awaitable property that returns the value of the action. It will raise an exception if the action failed.

await db.find(User.name == "Alice").count().value

Value Or

Value or is a method that takes a default value and returns the value of the action or the default value if the action failed.

count = await db.find(User.name == "Alice").count().value_or(0)

Raise on Errors

Raise on errors is a method that will raise an exception if the action failed. If the action succeeds it'll return nothing. It's a convenience method that allows you to raise errors and discard the result on success.

await db.find(User.name == "Alice").delete().raise_on_errors()

Database Results

DatabaseResult is a result type that is used to wrap the values of database actions. It provides a Success and Failure type that can be used to match on the result of an action.

match await db.find(User.name == "Alice").count():
    case DatabaseResult.Success(value):
        print(value)

    case DatabaseResult.Failure(error):
        print(error)

Value and Value Or

DatabaseResult objects provide a value property that returns the value of the action or raises an exception if the action failed. It also provides a value_or that takes a default value that is returned if the action failed.

Unlike AsyncResultWrapper, DatabaseResult objects do not need to be awaited.

result = await db.find(User.name == "Alice").count()
print(result.value)  # Raises an exception if the action failed

result = await db.find(User.name == "Alice").count()
print(result.value_or(0))  # Prints 0 if the action failed

Error

DatabaseResult objects provide an error property that returns the exception that caused the action to fail or None if the action succeeded.

result = await db.find(User.name == "Alice").count()
print(result.error)  # Prints None if the action succeeded

Lazy Loaded Relationships

Ommi provides support for lazy loading relationships between models. It fully supports forward references as string annotations. There are two supported relationship types using the ommi.query_fields.LazyLoadTheRelated and ommi.query_fields.LazyLoadEveryRelated generic types as annotations. It relies on models using the ommi.field_metadata.ReferenceTo annotation to define which field references which field on another model.

@ommi_model
@dataclass
class User:
    id: int

    posts: LazyLoadEveryRelated["Post"]

class Post:
    id: int
    author_id: Annotated[int, ReferenceTo(User)]

    author: LazyLoadTheRelated[User]

LazyLoadTheRelated and LazyLoadEveryRelated are awaitable to fetch the related models. They also provide an awaitable value property that returns the same value as well as a get method that takes a default value in case of a failure. LazyLoadTheRelated will return a single model while LazyLoadEveryRelated will return a list of models.

user = User(id=1)
posts = await user.posts

Lazy fields will only fetch once and then cache the result.

Get

'LazyQueryFieldsprovide aget` method that takes a default value and returns the value of the relationship or the default value if there is an error.

user = User(id=1)
posts = await user.posts.get([])

Value

LazyQueryFields provide a value property that returns the value of the relationship or raises an exception if there is an error.

user = User(id=1)
posts = await user.posts.value

Refresh

LazyQueryFields provide a refresh method that will fetch the related models again and update the cache.

user = User(id=1)
await user.posts.refresh()

Refresh if needed

The refresh_if_needed method will only fetch the related models if they haven't been fetched yet.

user = User(id=1)
await user.posts.refresh_if_needed()

Result

LazyQueryFields provide a result property that returns the tramp.results.Result object directly. This can be helpful for handling errors more explicitly.

user = User(id=1)
match await user.posts.result:
    case Value(posts):
        ...

    case Error(error):
        ...